Scholarlink

Proper speech helps preschoolers’ cognitive and social development. A hybrid model including Wav2Vec 2.0 for audio feature extraction, Bi-LSTM, and an attention mechanism improves preschoolers’ speech competency categorization. LSTM, Bi-LSTM with Attention Mechanism, and Proposed Bi-LSTM were examined. All tests confirm the Proposed model’s 98.26% accuracy, precision, recall, and F1 score. Confusion matrix: Ordinary 98.55%, Incompetent 100%, Proficient 96.0%. The recommended model has a 0.98 ROC AUC, compared to 0.95 for Bi-LSTM and 0.90 for LSTM, suggesting excellent class difference. Adding attention processes to Bi-LSTM enhances toddler linguistic proficiency. For the Proposed approach, speech delivery classification efficiency increases early childhood speech therapy.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Framework Using Feature Maps Improving Speech

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Arindam Bindlish,
  • Jitendra Singh Goyal,
  • Vaibhav Gujral,
  • Tanu Dua,
  • Gaura Dubey

摘要

Proper speech helps preschoolers’ cognitive and social development. A hybrid model including Wav2Vec 2.0 for audio feature extraction, Bi-LSTM, and an attention mechanism improves preschoolers’ speech competency categorization. LSTM, Bi-LSTM with Attention Mechanism, and Proposed Bi-LSTM were examined. All tests confirm the Proposed model’s 98.26% accuracy, precision, recall, and F1 score. Confusion matrix: Ordinary 98.55%, Incompetent 100%, Proficient 96.0%. The recommended model has a 0.98 ROC AUC, compared to 0.95 for Bi-LSTM and 0.90 for LSTM, suggesting excellent class difference. Adding attention processes to Bi-LSTM enhances toddler linguistic proficiency. For the Proposed approach, speech delivery classification efficiency increases early childhood speech therapy.